linux-kernel - [RFC PATCH 0/7] cross rename

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <1380643239-16060-1-git-send-email-miklos@szeredi.hu>
Date:	Tue,  1 Oct 2013 18:00:32 +0200
From:	Miklos Szeredi <miklos@...redi.hu>
To:	linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org
Cc:	viro@...IV.linux.org.uk, torvalds@...ux-foundation.org,
	hch@...radead.org, akpm@...ux-foundation.org, dhowells@...hat.com,
	zab@...hat.com, jack@...e.cz, tytso@....edu, mszeredi@...e.cz
Subject: [RFC PATCH 0/7] cross rename

This series adds a new syscall, renameat2(), which is the same as renameat() but
with a flags argument.  Internally i_op->reaname2() is also added, which can
later be merged with ->rename() but is kept separately for now, since this would
just blow up this patch without helping review.

The purpose of extending rename is to add cross-rename, a symmetric variant of
rename, which exchanges the two files.  This allows interesting things, which
were not possible before, for example atomically replacing a directory tree with
a symlink, etc...

The other reason to introduce this is for whiteout handling in union/overlay
solutions in an atomic manner without having to add complex code to each
filesystem's rmdir, mkdir and rename just for handling whiteouts.  With
cross-rename it becomes possible to handle whiteouts in a generic manner in most
of the cases:

rmdir(P) where P needs to be whiteout:
  - create-whiteout(orphan/X)
  - exchange(orphan/X, P)
  - rmdir(orphan/X)

mkdir(P) where P is currently a whiteout:
  - mkdir(orphan/X)
  - exchange(orphan/X, P)
  - unlink(orphan/X)

rename(P, Q) where P needs to be whiteout and Q is negative:
  - create-whiteout(Q)
  - exchange(P, Q)

rename(P, Q) where Q is a directory containing only whiteouts:
  - mkdir(orphan/X) (marked as opaque)
  - exchange(orphan/X, Q)
  - recursive-remove(orphan/X)
  - rename(P, Q)

The case that cannot be handled with cross rename is:

rename(P, Q) where P needs to be whiteout and Q exists

For this case a new rename flag will be propsed that atomically creates the
whiteout.

Thanks,
Miklos

---
Miklos Szeredi (7):
      vfs: rename: move d_move() up
      vfs: rename: use common code for dir and non-dir
      vfs: add renameat2 syscall and cross-rename
      ext4: rename: create ext4_renament structure for local vars
      ext4: rename: move EMLINK check up
      ext4: rename: split out helper functions
      ext4: add cross rename support

---
 arch/x86/syscalls/syscall_64.tbl |   1 +
 fs/dcache.c                      |  46 ++++-
 fs/ext4/namei.c                  | 379 +++++++++++++++++++++++++--------------
 fs/namei.c                       | 218 ++++++++++++----------
 include/linux/dcache.h           |   1 +
 include/linux/fs.h               |   2 +
 include/uapi/linux/fs.h          |   2 +
 7 files changed, 415 insertions(+), 234 deletions(-)

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/